Authors: Adam Rohde and Ashley Chiu
Date: December 6, 2019
Description: We deploy an experimental approach to hyperparameter tuning in the context of AlexNet. We investigate AlexNet’s innovations, amongst other hyperparameters known to influence accuracy, to screen for the most important hyperparameters in an AlexNet-type CNN.
Dataset: MNIST
Files:
-
Report: screening-alexnet-hyperparameters-with-mnist.pdf; screening-alexnet-hyperparameters-with-mnist.tex
-
Python Code: AlexNetExperiment.py (Note only one version of experiment code is provided here.)
-
R Code: AlexNetAnalysis.R
-
Experiment Data: AlexNetExperimentData_Iter1_1.csv; AlexNetExperimentData_Iter1_2.csv; AlexNetExperimentData_Iter2_1.csv; AlexNetExperimentData_Iter2_2.csv (Note that each author generated half the runs for each iteration.)